Abstract Intermittent or acute/cyclic hypoxia in tumors, with frequencies between a few cycles per minute to hours, is receiving increased attention because this type of hypoxia has been reported to have an influence on tumor malignancy as well as treatment resistance via increased expression of pro-survival pathways. Fast oxygen imaging methods are needed for the measurement of acute tumor hypoxia. Pulse electron paramagnetic resonance imaging (pEPRI) is a promising tool to provide three-dimensional partial oxygen pressure (pO2) maps in live tissues and tumors to assist with advanced studies of tumor biology, perfusion, drug development, and radiation treatment. Single point imaging (SPI), an acquisition technique developed at National Cancer Institute (NCI), is a subset of pEPRI methods that can be used for oxygen image acquisition. It provides high-resolution, high-fidelity images but is slow due to the need for acquiring each k-space point individually. In the current project, our goal is to improve the image acquisition speed of SPI by utilizing a combination of advanced hardware and deep learning. This will improve the imaging speed by many folds without compromising the image quality. These advances will be tested in a mouse model of fibrosarcoma tumor. This project will bring an NIH-developed technology to the commercial level. Our long-term goal is to imply the advanced hardware and software technologies of oxygen imaging to clinics to assist with oxygen-guided tumor treatments.